Author: Tom Likens
Date: 10:57:23 12/01/03
Go up one level in this thread
On December 01, 2003 at 08:57:55, Sune Fischer wrote:
>On December 01, 2003 at 08:19:50, Tord Romstad wrote:
>
>>On December 01, 2003 at 07:33:51, Tom Likens wrote:
>>
>>>I'm toying with adding full attack tables to my engine. Do you incrementally
>>>update your attack tables or do you generate them from scratch?
>>
>>It's funny that you ask this question right now, because I posted my new
>>attack table generation code just a couple of hours ago. :-) I don't want
>>to show you the old code, partly because it is too embarassing to show,
>>and partly because it is so mind-bogglingly ugly that the memory of having
>>seen it would probably give you nightmares for the rest of your life.
>>
>>I have always generated the attack tables from scratch at every node. The
>>complexity of updating them incrementally scares me, and those who have tried
>>usually claim that it is slower than generating them from scratch anyway.
>
>That will probably depend on the design of the table.
>
>I stored attack to and attack from information which can be used to do
>incremental updates.
>
>When moving a piece you take all the sliding pieces attacking the affected
>squares (that's just a few masks with bitboards) and invalidate their attack
>boards along with the moving piece and the one it might have captured.
>
>Now you have a list of pieces who's attack boards may be invalid.
>You run through this list and generate their new attack boards, you xor the new
>attack board with their old attack board (which is stored in the attack table)
>to get the squares which changed attack status, these you update.
>
>Hope that made sense :)
>
>I posted the code once, here is the main loop
>
>inline void BOARD::GenAttackers(uint32 affected,uint co) {
> register BITBOARD xors;
> register uint ix;
>
> while (affected) {
> ix = FirstBit32(affected);
> xors = attack_ix[ix];
> attack_ix[ix] = GenAttack(ty_ix[ix],sq_ix[ix],co);
> xors ^= attack_ix[ix];
> while (xors)
> attack_sq[GetFirstBitAndClear64(xors)] ^=(1<<ix);
> affected &= affected-1;
> }
>}
>
>affected is a 32 bit piece list
>attack_ix stores the attack board of piece with index ix
>attack_sq stores attackers to sq
>
>-S.
>>Tord
Hello Sune,
This looks interesting. It seems to me that you could do the same thing with
simply bitboards, *but* I'm guessing you use unsigned ints because they are
faster (i.e. more cache friendly).
--tom
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.